We already do all this in createVideoElement

We no logner use that library and the "qr-scanner" altternative turned out to
be better, so remove it completely.

We now queue it and in the callback check if the scan is still active,
and go from there. This means we redraw all the time and detect mostly
in the background.

This results in worse perf with the old library since that isn't async, but
we can always revert if needed.

This allows the detection to be async and pasue execution between capturing
the image, analyzing it and then finally drawing it.

With both being async it's hard to follow what would happen if both run at the same
time, so just prevent it and serialize start/stopping instead.

so 'this' is the same

This lets the user decide what to do (like scrolling)

Instead handle everything throuh property updates.

Instead stop it form outside which gives us more control.

This way we can extend the message with more data in the future if needed.

In case we move it out of view and back again we want a second event. The user can filter those
out if needed.

We need to give feedback to the user, so better let the user decide.

We use it for user feedback now, so there is no good reason to allow disabling it.

This allows us to show feedback to the user by coloring the clip border.
Also in case there is no match we don't send an event.